As is well known, the use of non-volatile memories of the EEPROM type as data storage devices is expanding on account of a specific feature of such memories whereby the information contained in the individual memory cells can be modified electrically in use, both at the writing and the erasing thereof, in an independent selective manner. For the purpose, a selection transistor is provided in series with each cell.
Consistently with today's technology, such storage devices rate very high in terms of reliability.
Concurrently therewith, there exists a demand for high capacity (at least 256 Kb) storage devices, incorporating an ever larger number of cells per memory unit. In this direction, the technology of electronic semiconductors is moving toward ever larger integration scales, and an attendant reduction in component size. However, this involves a more critical control of their characteristics and increased difficulty to ensure an acceptable level of reliability for such advanced cells fabricated to a very large scale of integration (ULSI).
Typically, non-volatile memory cells are integrated in a substrate of a semiconductor material and laid into rows and columns to form a matrix of cells. The cell element wherein the information is stored is a MOS transistor of the floating gate type. The logic state, or level, of the cell is defined by the amount of charge contained in the floating gate of the transistor.
In particular, the amount of charge is altered in EEPROMs by causing charges to flow by tunnel effect (known as Fowler-Nordheim current) through a thin layer of silicon oxide, the so-called tunnel oxide, which intervenes between the substrate of semiconductor material and the floating gate of the transistor.
Of the various types of EEPROMs, those having two levels of polysilicon will be considered here. Their associated memory cells comprise each a floating gate of polysilicon overlying a layer of gate oxide which is formed on top of a silicon substrate, and a control gate, also of polysilicon, overlying the floating gate and being isolated from the latter by a so-called interpoly dielectric layer.
The selection transistor associated with a cell can be variously constructed; for example, it consists of either a single layer of polysilicon or, again, two superposed layers of polysilicon provided above a gate oxide, with an interpoly dielectric layer optionally interposed. Where no interpoly dielectric is provided, it becomes known as a short-circuited double polysilicon construction. If the interpoly dielectric is provided, a short-circuit must anyhow be established between the two layers in appropriate areas of the device, either outside or inside the memory matrix.
In addition, circuitry transistors are usually provided in the device which may have a single polysilicon or double polysilicon with no interpoly dielectric layer. In particular, the last-mentioned instance will be considered in this specification by way of example.
The aforementioned storage structures can be fabricated with technologies that provide for alignment of the control gate to the floating gate. Examples of such processes can be found in U.S. Pat. No. 4,719,184 and European Patent EP-0255159.
One widely employed structure of EEPROM cells with a double level of polysilicon, known as FLOTOX, provides for the floating gate to be extended sideways to the tunnel oxide region outside.
A single memory cell of the FLOTOX type with a double level of polysilicon formed by a standard technique is described in U.S. Pat. Nos. 5,793,673 and 5,792,670, for example. A cross-sectional view of this is given in FIG. 1.
By way of example only, the cell is made with CMOS technology.
Referring to this Figure, the memory cell, generally referenced 1, comprises a series of a floating gate transistor 2 and a selection transistor 3, both of the MOS type, represented in an active area region of a substrate 4 having a first type of conductivity. The cross-section is taken in a parallel direction to the matrix columns, in particular along a source/drain line of the matrix of cells.
The floating gate transistor includes a tunnel oxide region 5 surrounded by a layer 6 of gate oxide. An upper first layer 7 of polysilicon forming the floating gate, and a second layer 8 of polysilicon used to form the control gate, are isolated electrically by an interpoly dielectric layer 9. The latter is usually a triple layer of oxide/nitride/oxide, or ONO, optionally overlaid with a thin layer of polysilicon, the so-called "polino". A low-resistivity silicide layer, not shown in the Figure, is usually provided on top of the second polysilicon layer 8.
The selection transistor 3 similarly includes a gate dielectric 11 beneath the first and second polysilicon layers 7 and 8, with the interpoly dielectric layer 9 extending therebetween.
Source and drain active regions of both transistors, having a second type of conductivity, are shown at 12, 13 and 14. In the storage transistor, they define a channel region 15 that includes specifically a tunnel area 15' under the tunnel oxide region 5. These memories further requires a doped continuity region 16 with a second type of conductivity which is typically formed by implantation and known as the capacitor implant, and which extends from the tunnel area to over the drain region 13 to provide electrical continuity between the substrate region lying beneath the tunnel oxide (tunnel area 15') and the drain region 13 of the cell, in operation.
As can be seen in the Figure, the two polysilicon layers are aligned along the source/drain direction in both the floating gate transistor 2 structure and that of the selection transistor 3.
Referring to FIGS. 2-5, some steps are illustrated therein of a CMOS process adopting a so-called DPCC (short-circuited double poly) flow for making a memory cell of the EEPROM type and a transistor of the external circuitry to the matrix, as is known in the art. It should be noted that both high-voltage or HV transistors, i.e., transistors capable of sustaining high voltages, and low-voltage or LV transistors, i.e., transistors operated at relatively low voltages, are typically formed in the circuitry. They distinguish themselves by the thickness of their gate oxide layer. However, a generic transistor, representing either an LV or a HV transistor, is shown by way of example in the drawings.
The drawings are cross-sectional views taken along the same section line as indicated in FIG. 1.
After defining the active areas, the substrate 4 is grown gate oxide layers 17 and 11 thereon, in the regions of the circuitry 20 and selection 3 transistors, and is grown a gate oxide layer 6 at the region of the storage transistor 2. Also defined is the tunnel oxide region 5 at the floating gate to be formed. The first polysilicon layer 7 is deposited and doped by implantation (FIG. 2). It is then partially defined by means of a mask having openings in the region 2 which lie in planes outside that of the drawings and are not represented.
The interpoly dielectric layer 9 and a thin polysilicon layer 21 are then deposited. The last-mentioned layer provides protection for the underlying interpoly dielectric during the next masking step. At this stage, as shown in FIG. 3, a selective etching step is carried out to remove the thin polysilicon layer 21 and the underlying interpoly dielectric layer 9, in those regions of the storage device where the external circuitry is to be formed, using a mask 22 which leaves the portion intended for the cell matrix covered. More generally, this mask covers the device regions which are intended to accommodate circuit elements with a DPCC structure.
The following step, not illustrated, includes implanting the transistors in the circuitry to define their threshold, the so-called LVS (Low Voltage Shift) implant, e.g., by implantation of a dopant such as boron. This is a low-energy, light-dosage implantation. No mask is needed at the matrix because the interpoly dielectric layer provided therein will screen off the low-energy implant, specifically by the presence of nitride.
Thereafter, the second polysilicon layer 8 and a silicide layer 23 are deposited. A selective etching step is then carried out using the mask 24. During this step, illustrated by FIG. 4, the silicide layer 23, second polysilicon layer 8, and thin polysilicon layer 21 in the matrix are removed, while from the circuitry, only the silicide layer 23 and the second and first polysilicon layers 7 and 8 are removed, there being no interpoly dielectric layer present.
At this stage (FIG. 5), without removing the previous mask 24, a self-aligned etching step is carried out to etch away the intermediate dielectric layer 9 and the first polysilicon layer 7 in the matrix, using a double-resist mask 25. Thus, the cell structure, including the selection transistor, is completely defined in a self-aligned manner. The mask 25 fully exposes the matrix region where the two polysilicon layers are not short-circuited, while completely masking the circuitry region defined during the previous step.
In an advanced process like that just described, whereby the overall size of the memory structure is greatly reduced, a critical factor in current technologies is control of the threshold value of the data storage transistor. In fabricating integrated circuits with methods of the CMOS type, whereto the present invention specifically relates, it is particularly difficult to establish a low threshold voltage with a small gap between the drain and source active regions.
It should be recalled in this respect that non-volatile memories of a standard type are programmed in either of two logic states, namely written and erased, or "0" and "1", to store one bit per cell. A different level of the threshold voltage corresponds to each of them. The cell programming, i.e., the storing of electric charges in the floating gate region, depresses the cell threshold level. Subsequent erasing of the cell brings its threshold back to its former value. The state of a virgin cell corresponds to a voltage level intermediate those of an erased cell and a written cell.
The threshold values are determined by the cell structure, as well as by construction features and properties of the materials comprising it. In a standard memory cell, reading the threshold value of a virgin cell may cause trouble at the programming stage. In particular, the problem is one of being able to set the threshold value of a virgin cell such that it can be read as if it were a written cell or an erased cell, according to the specific requirements of the device which is to include it.